Presentation
2019 IEEE Sidney Fernbach Award
Presenter
Event Type
Awards Presentation
TP
TimeWednesday, 20 November 20199:30am - 10am
LocationMile High Ballroom
DescriptionHow the Julia Language Can Fulfill the Promise of Supercomputing
During the first phase of HPC, the emphasis has been on performance above all else. The thinking that carried the day was that you can hack hard enough, bring up performance, and earn your brownie points. The Julia language, through abstractions and multiple dispatch technology, is making it possible to build supercomputing applications more readily in ways that can manage the complexities of modern supercomputers.
We will explain how Julia is transforming HPC in such areas as Scientific Machine Learning, Climate Science, and Medicine. We will conclude with the notion that the one true metric for HPC should not be performance; it should be the number of users. Performance, productivity, scalability, reproducibility, composability and other obvious and non-obvious metrics can then be subsumed by this “prime directive”.
During the first phase of HPC, the emphasis has been on performance above all else. The thinking that carried the day was that you can hack hard enough, bring up performance, and earn your brownie points. The Julia language, through abstractions and multiple dispatch technology, is making it possible to build supercomputing applications more readily in ways that can manage the complexities of modern supercomputers.
We will explain how Julia is transforming HPC in such areas as Scientific Machine Learning, Climate Science, and Medicine. We will conclude with the notion that the one true metric for HPC should not be performance; it should be the number of users. Performance, productivity, scalability, reproducibility, composability and other obvious and non-obvious metrics can then be subsumed by this “prime directive”.